go network_error ai_generated true

rpc error: code = Unavailable desc = all SubConns are in TransientFailure

ID: go/grpc-unavailable-load-balancer

Also available as: JSON · Markdown · 中文
80%Fix Rate
87%Confidence
0Evidence
2024-08-10First Seen

Version Compatibility

VersionStatusIntroducedDeprecatedNotes
1.21 active — — —
1.22 active — — —

Root Cause

The gRPC client's load balancer cannot connect to any backend server because all endpoints are unreachable or failing health checks.

generic

中文

gRPC客户端的负载均衡器无法连接到任何后端服务器，因为所有端点都不可达或健康检查失败。

Workarounds

  1. 85% success Verify server health and restart any down instances. 
    // Check server status via health check endpoint
healthClient := grpc_health_v1.NewHealthClient(conn)
resp, err := healthClient.Check(ctx, &grpc_health_v1.HealthCheckRequest{Service: "myservice.MyService"})
  2. 90% success Use a different load balancing policy like round_robin with proper service discovery. 
    conn, _ := grpc.Dial("dns:///myservice:50051", grpc.WithDefaultServiceConfig(`{"loadBalancingPolicy":"round_robin"}`))

Dead Ends

Common approaches that don't work:

  1. Recreating the client without changing server addresses. 90% fail

    Same endpoints, same connectivity issue.

  2. Setting a shorter dial timeout. 80% fail

    Timeouts don't fix unreachable servers.